I checked my GRE Biology scores at 12:08 am this morning, after awaiting them for the past week and being blocked off the score report page all day Sunday by the “routine maintenance” ETS was undertaking on everyone’s accounts. I don’t mean to brag, but I was very pleased with my results:
In preparation for the real thing, I had taken the official practice test offered by ETS twice: first about a month before my exam date, before I had started studying, and second the evening before the exam, to gauge what I had learned. For comparison, that second time, I scored a 910 total (98th percentile), with an 85 (91st), 91 (97th), and 92 (99th) in the cell/molecular, organismal, and ecology/evolution sections respectively. These numbers were the benchmarks I’d set for myself going into the test at 8:30 am on October 24th.
Thus, my real exam scores reflect improvements almost across the board from the night before to the day of the test, in all categories except ecology/evolution. I attribute much of this to the last minute studying I did the night before the test, sitting in bed going through my wrong practice test answers and mastering concepts that I was supposed to have finished weeks before. However, I also believe a big part was starting the real test in medias res, by beginning with the data analysis problems (see my previous GRE post for further details).
This would also explain my unfortunate drop in the ecology/evolution section. By the new test-taking scheme, most of the questions I had saved for last were in the ecology/evolution recall section. Since these were my final questions and I was pressed for time, I definitely got sloppy, reading too quickly, not thinking thoroughly, and filling in answers for questions that were best left blank. It also didn’t help that I believed ecology/evolution to be my ‘easy’ section: On my practice test, the same topic saw me score in the 99th percentile both times, so I thought I could get away with it again without reviewing some concepts that were huge on the real test (looking at you, ecosystem ecology and equilibrium theory of island biogeography).
Overall, however, I am extremely satisfied with my scores on this GRE subject test. Although I came up slightly short in my strongest subject, the difference was insubstantial, especially considering my marked improvement in my two weaker subjects. By basically replicating my performance on the practice test, I did what I set out to do, and got the best results I could hope to use to top off my applications.
Survey a group of intro bio freshmen on what factors drive evolution, and natural selection (and its close allies, artificial and sexual selection) will assuredly be almost an exclusive answer. This is understandable, because Darwinian selection produces the most spectacular and most predictable changes in populations over time, from anatomy to physiology to behavior to development. But besides Darwinian selection, other forces also drive the evolution of populations: random drift of allele frequencies over generations, migration between neighboring populations, and mutation of genetic material itself. But these forces are so difficult to predict or observe that even biologists have a relatively poor understanding about these compared to Darwinian selection.
Given the delicate gene networks involved in regulating cell growth in the human body, our constant exposure to cancer-causing environmental agents, and just the sheer number of cells that are born and die over the course of a human lifetime, potentially cancerous cell lineages probably arise very often, but are almost always nipped in the bud by our immune system. The few cells that escape detection and establish tumors are also susceptible to medical treatments, including chemotherapy and outright removal. So it should make sense that mutated cells that do lead to cancer must have had some heritable traits that allowed them to grow rapidly and survive persecution and were passed down to their descendants: That is, cancer cells must evolve by Darwinian selection. But a new paper by Ling et al. out in PNAS asserts that this intuitive assumption might be a poor description of cancer evolution in the real world.
As the authors point out, simple math would predict a high amount of genetic diversity generated by mutation within a single tumor, which is essentially a big population of cancer cells. Though the expected mutation rate is only about 1% of the (functional) genome mutated per cancer cell division, an average tumor can grow to a population size of millions or billions of cells, so multiple mutations ought to occur during a tumor’s developmental history. If there is a strong selective force experienced by this population that favors certain cancer cell genotypes over others, one would expect very few of these mutations to survive; instead, the tumor would contain a homogenous genetic profile after the extinction of less beneficial mutations. Indeed, previous studies that sampled tumors found that each contains only tens to hundreds of distinct genotypes, so scientists simply assumed that Darwinian selection was taking place among cancer cells.
To test whether these estimates were accurate, Ling et al. analyzed a single liver tumor, just 3.5 cm in diameter, at high resolution. From 286 samples punched out of a mere 1 mm thick slice from the middle of the tumor, they verified and classified 269 mutations called single-nucleotide variations (SNVs) (other mutation types, based on abnormal numbers of gene copies rather than gene alteration, were ignored for logistical reasons). From these numbers, the authors determined that about 100 million coding SNVs would have occurred during the tumor’s growth into a population of billions of cells.
After sorting out which samples contained which SNVs, the authors were able to construct a full phylogeny, a family tree, to illustrate the history of diversification between all the samples. Interrelated samples were identified by shared SNVs, and major branches of related samples were classified into 20 families called clones, delineated by the SNVs unique to that branch. Interestingly, genetic relationships between the samples were also reflected in the structure of the tumor, in that closely related samples were located close together and nested within the same clone, and younger, more mutated clones were located closer to the edge.
If Darwinian selection was occurring within the tumor, one would expect certain clones to have grown faster, and thus be more prevalent, than other clones. Mutated genotypes that confer the greatest overall advantage on a cancer cell’s growth or survival would be swept to fixation in the population, crowding out cells unlucky enough not to carry those mutations. This was not the case, as the clone sizes instead corresponded closely to a null model in which selection was entirely absent. Indeed, the population contained many, many SNVs that each occurred within only a few cells in the tumor, as if mutations were generated rapidly but, being neutral in their effects on cell fitness, were never extinguished. Even the shape of the tumor and spatial distribution of the clones suggests that no cell lineages grew and divided outward at a faster or slower rate than others.
I can’t pretend to be an expert on this topic or the techniques and models the authors used, but I found the conclusions of this study to be quite robust, even conservative. Remember that their conclusions rest on characterizing a single type of mutation and searching for them only in the coding portion of the genome (that is, the 2% of the genome that code for proteins and so are directly responsible for a cell’s characteristics). If they were to have complicated their models by including the prevalent gene duplication events, or expanded their mutation screening to the 98% of DNA that doesn’t code for anything, it’s possible that they would have reached the same conclusions. In the future, I’d like more direct evidence for genetic diversification within a tumor as it grows, to complement evidence taken after the fact. Of course, this opens a whole can of worms on whether lineages of HeLa cells grown in vitro for innumerable generations are equivalent to tumors from a human or animal model.
Cancer is a broad category of diseases that develop in different ways and are caused by many unknown factors, and variables such as tumor structure can change how strongly Darwinian selection influences cancer cell evolution. Ultimately, the authors take a page from the rest of evolutionary biology, and suggest that future studies assume a null model of no Darwinian selectionin order to evaluate hypotheses about selective forces in cancer. Clearly, the study of evolutionary processes in cancer, and how they shape disease trajectories and patient outcomes, is still in its early stages.